Real Estate Valuation#

Market data of house sales in the Sindian District of New Taipei City, Taiwan, with the house price of unit area to be predicted from the age and the surroundings of the house. It is a small regression dataset: 414 samples and 6 numeric features without missing values, including the geographic coordinates of the houses.

Overview#

Loader

load_real_estate_valuation()

Task

Regression

Samples

414

Features

6

Targets

1 continuous target, house_price_of_unit_area

Missing values

None

License

CC BY 4.0

Features#

All six features are numbers, stored as float64, and none of them is categorical:

  • transaction_date, the date of the sale as a number: 2013.250 is March 2013 and 2013.500 is June 2013. Only twelve different dates occur, from 2012.667 to 2013.583.

  • house_age, the age of the house in years.

  • distance_to_nearest_mrt_station, the distance to the nearest MRT station in meters.

  • number_of_convenience_stores, the number of convenience stores in the living circle on foot. It is a whole number from 0 to 10.

  • latitude and longitude, the geographic coordinates of the house in degrees.

The dataset has no missing values. The identifier column No of the source is not stored.

The dates are stored as ucimlrepo gives them, rounded to three decimals. The xlsx file of the UCI page has seven decimals (2012.9166667 for the first sale), so the stored dates differ from it by 0.0004 at most.

Target#

y is the house price of unit area, in 10000 New Taiwan Dollar per Ping, where a Ping is a local unit of area, 1 Ping = 3.3 square meters. In the data file of the source it goes from 7.6 to 117.5.

The source says that the data was randomly split into a training part (2/3 of the samples) and a testing part (1/3), but its file does not tell which sample belongs to which part. The dataset holds all 414 samples and does not provide the split, so make your own, for example a random split with a fixed seed.

Source and citation#

Please cite the dataset as:

Yeh, I. (2018). Real Estate Valuation [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5J30W

See also:

  • Yeh, I., & Hsu, T.-K. (2018). Building real estate valuation models with comparative approach through case-based reasoning. Applied Soft Computing.

License#

CC BY 4.0, as stated by the source. The license was checked at the source on 2026-11-09. See License and citation.

Changes made#

  • Converted to the NumPy NPZ format. The features and the target are stored as float64.

  • The identifier column No, the numbers 1 to 414, is not stored, because it is not a feature.

  • The column names are written in lowercase with underscores, and the prefixes X1 to X6 and Y are dropped, for example X3 distance to the nearest MRT station becomes distance_to_nearest_mrt_station.

Note

The source says that the data set was randomly split into a training part (2/3 of the samples) and a testing part (1/3), but its file has no column that tells which sample belongs to which part. All 414 samples are stored and the split is not provided.

Note

The transaction date is a number: 2013.250 is March 2013 and 2013.500 is June 2013, as the source describes. Only 12 different dates occur, from 2012.667 to 2013.583. ucimlrepo gives the dates rounded to three decimals, while the xlsx file of the UCI page has seven (the first sale: 2012.9166667), so the stored dates differ from that file by 0.0004 at most.

Note

The target is the house price of unit area in 10000 New Taiwan Dollar per Ping, where a Ping is a local unit of area, 1 Ping = 3.3 square meters, as the source gives.

Example#

>>> from pyml_datasets import load_real_estate_valuation
>>> data = load_real_estate_valuation()
>>> data.X.shape, data.y.shape
((414, 6), (414,))
>>> data.has_missing
False
>>> data.feature_names[2]
'distance_to_nearest_mrt_station'
>>> data.target_names
('house_price_of_unit_area',)